Wasserstein rates of all orders for the Sato-Tate law of Kloosterman sums
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Mathematics. The day in mathematics: new theorems, new proofs, and what they open.
Ada: We're looking at the paper: "Wasserstein rates of all orders for the Sato-Tate law of Kloosterman sums".
Emmy: Here's the gist of the paper. The paper establishes upper bounds for Wasserstein rates between the distribution of normalized Kloosterman sums and the Sato-Tate law, providing quantitative equidistribution results that are significant in number theory The bound Ws(µq, µST) is shown to be bounded by specific functions of q and s for various ranges of s.
Ada: Main Theorem Statement Theorem A provides the main result for the Wasserstein rates between the distribution of Kloosterman sums and the Sato-Tate law The bound Ws(µq, µST) is given by a piecewise function depending on s, specifically Ws(µq, µST) ≪ q − 1/3 for 1 ≤ s ≤ 2, q − 1/6−1/(3s) for 2 ≤ s < ∞, and q − 1/6 for s = ∞.
Emmy: This result answers a question posed by Kowalski and Untrau regarding bounds for Ws(µq, µST) when s is not equal to 1.
Ada: Theoretical Framework and Tools The proof relies heavily on the quantile coupling technique which arises from Stieltjes sums over two different intervals whose cells are paired so as to balance two measures This coupling is optimal for all s simultaneously, and it is used to establish bounds for Ws(µ, µST).
Paper summary: Emmy: So wrapping up this paper, "Wasserstein rates of all orders for the Sato-Tate law of Kloosterman sums", the authors have done a lot by quantifying how fast the distribution of Kloosterman sums converges to the Sato-Tate law.
Ada: They did that by using quantile coupling, which they say is optimal for all s simultaneously, and they tied this to explicit smoothing inequalities relative to the Sato-Tate law.
Kolya: It’s interesting because numerically, for primes in the range of one hundred three to six million, they found that all these distances seemed to decay like p minus one over two.
Emmy: And that numerical observation leads them to conjecture a general bound: they suggest the rate is bounded by s minus epsilon p minus one over two plus epsilon for any s greater than or equal to one.
Ada: So, what this really means for us is that we have more precise control over the error when studying these arithmetic sums. It moves the discussion from just saying "they are close" to saying "they are close at a rate of this specific order."
Emmy: It gives us a better structural understanding of the relationship between these two distributions, even though the proofs rely on tools like W2 smoothing inequalities and comparisons to Kolmogorov distance.
Kolya: And one limitation they state is that their results for W2 depend on the Chebyshev moments of the distribution, and they show that if those moments are bounded by Katz type bounds, then these Wasserstein bounds hold.
Ada: So the paper sets a very clear benchmark for what we expect this kind of error to look like in number theory. It’s a foundation for understanding how well these sums distribute themselves across the space defined by the Sato-Tate law.
Conclusion: Ada: So we've been looking at how Kloosterman sums distribute themselves against the Sato-Tate law, and this paper is titled "Wasserstein rates of all orders for the Sato-Tate law of Kloosterman sums."
Emmy: Yeah, it’s about getting really precise about that convergence. The authors are trying to give us quantitative bounds on how fast these two distributions get closer.
Ada: Exactly. They’re not just saying they're close; they’re giving us a formula for the error rate, or the Wasserstein rate, for different orders of closeness.
Kolya: From a probabilistic standpoint, what this means is that we have a predictable tail behavior for these sums relative to the Sato-Tate distribution. It tells us how likely it is to get a large deviation from that expected structure.
Emmy: Structurally, I think it shows how the underlying geometry of these sums dictates exactly how fast they settle down into the Sato-Tate shape. It’s about finding that structural relationship you can map out with these couplings and inequalities.
Ada: And numerically, they found some decay rates that look really clean for many primes. It seems like the convergence happens at a rate of something quite specific, depending on how far out we go in those sums.
Kolya: That's what I find interesting—the numerical data supports a conjecture they’re making about the general behavior across all orders. It suggests that this decay isn't just an artifact of the specific primes they tested; it seems to be a more fundamental property.
Emmy: It opens up questions about where these bounds hold and what conditions we need for those specific rates to actually apply when you move from theory to application in number theory.
Ada: Right. So, while they give us some very concrete bounds, the big question now is how these rates connect to other areas of number theory we’re studying.
math.NT, math.PR
Submitted: 2026-10-12
Updated: 2026-11-03
Comments: 18 pages. Substantially revised version
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 93/100
The gist: The paper establishes upper bounds for Wasserstein rates between the distribution of normalized Kloosterman sums and the Sato-Tate law…
Key concepts
- Wasserstein rates (Ws)
- These rates measure the distance between two probability distributions. In this paper, it quantifies how far the distribution of normalized Kloosterman sums is from the Sato-Tate law using different orders of distance, like W2 or Ws.
- Sato-Tate law (µST)
- This is a fundamental law in number theory describing the statistical distribution of certain arithmetic objects, specifically related to the eigenvalues of modular forms. It serves as the target distribution that Kloosterman sums are being compared against.
- Quantile coupling technique
- This is a mathematical method used to prove bounds on Wasserstein distances. It involves pairing up intervals from two different distributions so that their measures balance each other out, which helps establish optimal coupling for measuring the distance between them.
Terminology
Summary
The paper establishes upper bounds for Wasserstein rates between the distribution of normalized Kloosterman sums and the Sato-Tate law, providing quantitative equidistribution results that are significant in number theory The bound Ws(µq, µST) is shown to be bounded by specific functions of q and s for various ranges of s. p. 0
Main Theorem Statement
Theorem A provides the main result for the Wasserstein rates between the distribution of Kloosterman sums and the Sato-Tate law The bound Ws(µq, µST) is given by a piecewise function depending on s, specifically Ws(µq, µST) ≪ q − 1/3 for 1 ≤ s ≤ 2, q − 1/6−1/(3s) for 2 ≤ s < ∞, and q − 1/6 for s = ∞. This result answers a question posed by Kowalski and Untrau regarding bounds for Ws(µq, µST) when s is not equal to 1. p. 0
Theoretical Framework and Tools
The proof relies heavily on the quantile coupling technique which arises from Stieltjes sums over two different intervals whose cells are paired so as to balance two measures This coupling is optimal for all s simultaneously, and it is used to establish bounds for Ws(µ, µST). The proof also utilizes an explicit W2 smoothing inequality relative to µST and a sharp comparison between Ws and the Kolmogorov distance for laws with square-root edges. p. 0
Bounds for Different Orders of Wasserstein Distance
The paper presents several bounds depending on the order s of the Wasserstein distance. For s = 2, the bound follows from a smoothing inequality of Borda and Cuenin on SU2. p. 0 For s > 2, a weaker inequality is given by Ws ≪ q − s/(6(s−1)) (Remark 6. p. 01). The bound for the infinite order Wasserstein distance W∞ is shown to be W∞(µq, µST) ≪ q − 1/6. p. 1
Numerical Observations and Conjecture
Numerically, for primes 103 < p < 10 6, all these distances appear to decay like p − 1/2. The conjecture suggests that Ws(µp, µST) is bounded by s − ε p − 1/2 + ε for every 1 ≤ s ≤ ∞ and ε > 0. p. 1 Furthermore, the computed values of Ws(µp, µST) decay like p − 1/2 for every s tested, including s = ∞. p. 1
Related Results and Further Exploration
Theorem B provides bounds for Ws(µ, µST) based on the discrepancy D(µ). For example, if D(µ) ≤ 1/8, then Ws(µ, µST) ≪ s − 3/(2s) for s > 3. p. 0 Theorem C provides bounds for the Kolmogorov distance D(µ), which is related to W2 via Theorem 5. p. 63. The paper also discusses the relationship between Ws and W∞, noting that for compactly supported measures on R, Ws(λ, ν) s = lim π→0 ordered balanced Sπ(x − y s). p. 2
Key Definitions and Lemmas
Definition 2.3 defines a paired partition as balanced if λ(Ii) = ν(Jσ(i)) for all i. p. 2 Lemma 4.2 provides bounds on the increments of the quantile function q, showing q(t + h) − q(t − h) ≤ Cht−1/3∗ if t∗ ≥ 2h. p. 2 Lemma 5.3 gives a bound for W2(µ, µST) using the Chebyshev moments. p. 0
Smoothing Inequalities and Measures on SU2
Lemma 5.1 establishes a smoothing inequality for W2(ρ, ν) based on the H˙ −1 norm. p. 1 Lemma 5.2 provides an estimate for the measure ηT on SU2, showing Rχk dηT = 0 for k > T. p. 5 Theorem 5.3 uses these results to bound W2(µ, µST) using a coupling with a central probability ηT. p. 0
Conclusion and Open Questions
The paper concludes by noting that for all computed primes, µcp(k) ≤ 0.31(k+1)p − 1/2 for k ≤ 40. It also poses several questions regarding the nature of W2 bounds, the closeness of µp to µST compared to random samples, and analogues for hyper-Kloosterman sums. p. 7 The numerical data suggests that W2(µp, µST) is significantly smaller than expected by some thresholds. p. 6
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved analysis of Kloosterman sum distributions for cryptographic applications by leveraging Wasserstein distance bounds like Ws(µq, µST) ≪ q(-1/3) for 1 <= s <= 2. This allows AI systems to more accurately estimate the deviation of actual Kloosterman sum distributions from the idealized Sato-Tate law, potentially enhancing the security analysis of protocols relying on these sums.
-
Enhanced estimation of discrepancy in finite field point sets by utilizing bounds derived from Theorem C, which states that D(µ) ≤ 19(Aδ) 1/2 for empirical measures with bounded Chebyshev moments. This enables AI systems to provide tighter guarantees on the uniformity and equidistribution properties of points generated within finite fields, improving the robustness of distributed algorithms.
-
Development of a more precise convergence rate analysis for exponential sums by applying Theorem A, which provides specific bounds like Ws(µq, µST) ≪ q(-1/6 - 1/(3s)) for s >= 2. This allows AI models to predict the asymptotic behavior of these sums with higher accuracy as the field size q increases, which is critical for analyzing pseudorandom number generators based on these sums.
-
Improved error estimation for empirical distributions by incorporating results from Theorem 5.3, which bounds W2(µ, µST) using Chebyshev moments: W2(µ, µST) ≤ 2 sqrt(C0/T + 4X T k=1 µb(k) / (2k 2)). This allows AI systems to quantify the error in approximating a complex probability distribution using finite samples, leading to more reliable statistical inference on data generated within these arithmetic settings.
-
Creation of a robust framework for comparing empirical point sets against ideal distributions by utilizing the quantile coupling results, specifically that for 1 <= s < infinity, Ws(λ, ν) s = lim π→0 π ordered balanced Sπ(x − y s) = lim η→0 inf π<η balanced Sπ(x − y s). This allows AI systems to establish a direct quantitative relationship between the empirical distribution of points and the true underlying target distribution, aiding in the validation of sampling methods.
Sources
- Smoothing inequalities for transport metrics in compact spaces
- Wasserstein metrics and quantitative equidistribution of exponential sums over finite fields
Related papers
- K ab has the Bogomolov property for canonical heights
- Quasi-Jacobi forms, Appell-Lerch functions, and false theta functions as q-brackets of functions on partitions
- Riemann functions for almost primes and their counting errors
- Topological flatness of orthogonal spin local models
- Asymptotic counting of integers with prime factors p r a s b
- Pellarin's Identity, Carlitz Period, and Anderson Generating Functions over Arbitrary Curves